Visual tracking based on biologically inspired C2 feature and whole constraint

Biologically inspired C2 feature has been introduced to visual tracking problem in the previous work. Because of the scale invariance and the complex shape selectivity of C2 feature, the tracking model can obtain good performance even in some challenging videos. However, the C2 feature lacks of the relative position relationship among the patches. It may cause tracking drift when dealing with approximately symmetric object, such as faces. A new visual tracking model is proposed based on biologically inspired C2 feature and whole constraint in this paper. The whole constraint is used to make an overall constraint of the patch position relationship. The experimental results show that the improvement is effective.